My use-case is the following:
A vector of size n read from a binary file.
Among other variants (iostreams, in my case custom code doing decompression), I can do something with semantics like this:
vector<myElem> v;
v.resize(n); // from my understanding v now has size n and capacity n
fread(v.data(), sizeof(myElem), n, myFile);
However, later I will have to (repeatedly) add and remove two elements to such a vector. (While this sounds pretty stupid, it can have positive effects to add sentinel values to lists so that intersections of sorted lists do not have to add comparisons for bound-checking).
For that, I would love to prealloacte a vector with size n and capacity n + 2.
I think I could do something like:
vector<myElem> v;
v.resize(n + 2); // from my understanding v now has size n + 2 and capacity n + 2
v.pop_back();
v.pop_back(); // v now has size n and most probably still has capacity 2 (unless n is pretty small)
fread(v.data(), sizeof(myElem), n, myFile);
Obviously, this is neither pretty nor guaranteed to behave as I would liek it to. In practice, I think it really should behave that way for big n and if small n should ever occur, a reallocation doesn't matter.
Still, it would be great to hear if there are better ways.
edit:
I am unsure how I can make use of reserve in my case. If I reserve a capacity of n + 2, the vector still has size 0. If I resize to n, i also change the capacity.
If I resize first and then reserv, I allocate memory two times and copy the whole vector in the process.
You can use v.reserve(n + 2) to change the vector's capacity without altering its size. Take a look at the documentation to better understand what is going on.